B40-12 Creation and Validation of an Automated Calculator for Primary Graft Dysfunction After Lung Transplant

Rationale

Primary graft dysfunction (PGD) remains an important predictor of short- and long-term lung transplant outcomes. While the criteria are well-defined, the most accurate assessment of PGD requires several data points that may not be available to the clinical team in real time (such as during clinical rounds) and is most accurate when tabulated retrospectively. The objective of the current study was to formulate a PGD calculator within our institutional data repository - Pulmonary Transplant Research Electronic Environment (Pulm-TREE) - to automatically and reliably grade PGD in lung transplant recipients and support research efforts.

Methods

Data extraction for Pulm-TREE analyses were conducted using an IRB-approved protocol for our institutional research registry. The Pulm-TREE PGD calculator synthesized and assessed an automated EHR data extraction of graft reperfusion times during surgery, radiographic assessment, extracorporeal membrane oxygenation utilization, ventilator settings, oxygen supplementation, arterial blood gases, and oxygen saturation values. The calculator interpreted radiographic reports for the presence of pulmonary edema and evaluated PGD scores at 0, 24, 48, and 72 hours post-lung transplant based on the 2016 International Society of Heart and Lung Transplantation consensus report. PGD scores provided by the calculator were compared to scores provided by an expert clinical adjudicator who was blinded to the calculator score. Analysis of agreement was assessed using descriptive statistics and Cohen’s Kappa.

Results

Using 117 evaluable time points from 30 randomly selected registry patients, the adjudicator was in agreement with the PGD status assigned by the calculator 96/117 instances (82.1%, Cohen’s Kappa = 0.74, p < 0.001). The adjudicator identified PGD grades as follows: PGD0 = 55 (47%), PGD1=19 (16.2%), PGD2=10 (8.5%), and PGD3=33 (28.2%). Disagreement (17.9%) was entirely attributable to differences in chest radiograph interpretation. The calculator identified pulmonary edema more frequently, accounting for 17/21 disagreements with the adjudicator.

Conclusion

A PGD calculator may be a useful addition to PGD evaluation and research, and has potential for integration within a clinical electronic medical record system. Further efforts to validate the calculator in a larger prospective cohort are ongoing.

This abstract is funded by: NIH